[antlr-interest] question about tree parsing (2.x)

[Lloyd Dupont]

Hi Thanks Brian (and Mounty for his good idea as well)
I do that because I used the a customized GNU-C parser to analyse ObjectiveC 
header
but then, in the tree parsing phase, only a limited subset of the produced 
tree interest me (my ObjectiveC class + typename & # of STAR in cast), so I 
have to parse part of existing tree and ignore the rest.

And I was wondering if these syntax of mine were ok to visit part of big 
tree.
But I'm not sure, I still have a few bug when I do that...

----- Original Message ----- 
From: "Bryan Ewbank" <ewbank at gmail.com>
To: <antlr-interest at antlr.org>
Sent: Friday, June 03, 2005 7:17 PM
Subject: Re: [antlr-interest] question about tree parsing (2.x)


Yes; you can do this; you are probably seeing ambiguity warnings about
the loop because both "ID4" and "." will match an ID4 node.  The trick
is a syntactic predicate.  Change this:

> bottomNode
>     :    #(
>             ID3
>             (
>                     ID4
>                 |   . // catch all for ID5 | ID6
>             ) // '+' volontary removed, exit should get 1 level up, no?
>         )
>     ;

To this:

> bottomNode
>     :    #(
>             ID3
>             (
>                 (ID4) =>  ID4
>                 |   . // catch all for ID5 | ID6
>             )+
>         )
>     ;

The "(ID4) => ID4" tells ANTLR to use that selection if an ID4 is seen.

I don't see why you removed the "+" unless you only want to match the
first child of ID3.  To match all ID4 children of ID3, you need the
"+".